Harness Autonomy Tiers (HAT)
Grant autonomy at a pace you can defend.
You do not have to grant full autonomy on day one. The harness is the product; your organization chooses the harness autonomy tier, or HAT, it starts at, a ladder from fully supervised to sovereign, and grows toward full autonomy as the trail earns trust. Each HAT grants more autonomy with more governance underneath. The audit substrate is the same at every HAT. What changes is how much the human is in the loop.
HAT 01
Supervised
Every action is proposed. A human approves.
The agent does the work and proposes the action. A human approves before anything executes. The substrate records the proposal, the decision, the approver, and the timestamp. Nothing happens without a name attached to it.
Governance
Human-in-the-loop on every action.
Who it is for
First deployments. Highest-scrutiny workflows. The tier where you start before the audit trail has earned trust.
HAT 02
Coordinated
Agents hand off to each other within policy.
Multiple agents coordinate across systems and steps, handing work to each other and to humans. Every handoff is attributed and recorded against the policy in force at the time. The unit of work becomes the swarm, and the swarm is fully traceable.
Governance
Cross-agent handoffs, governed at the moment of transfer.
Who it is for
Multi-domain operations under governance. Outage response, case assembly, review-in-flight, anywhere the work crosses more than one system.
HAT 03
Autonomous-within-Policy
Agents decide and act inside bounded policy.
The agent acts on its own authority, inside policy you wrote. Humans are in the loop on exceptions, not on every step. The boundary is the policy, the policy is code, and the audit chain proves the agent stayed inside it.
Governance
Bounded autonomy. Human escalation on exception only.
Who it is for
Scaled operations where the audit trail has earned trust and the cost of approving every step has become the bottleneck.
HAT 04
Air-gapped
The full substrate, with zero external dependency.
Every capability of the substrate, deployed with no call home, no external service, no substrate dependency outside your boundary. The substrate is sovereign by construction. Full sovereignty, where even the model behind the agent runs inside your boundary and nothing leaves at all, is the path the model-training pipeline opens.
Governance
Sovereign substrate deployment. The substrate has no external dependency.
Who it is for
Classified, regulated, and sovereignty-required environments. The tier for when you cannot delegate control to anyone, including us.
The throughline
One audit substrate. Four levels of trust.
The HATs are not four different products. They are one substrate, configured for how much autonomy your governance is ready to grant. A HAT 1 agent and a HAT 4 agent write to the same immutable, tamper-evident audit chain. The difference is where the human sits.
That matters because trust is earned, not declared. You can start every agent Supervised, watch the audit trail accumulate, and promote a workflow to the next HAT only when the evidence supports it. The substrate makes the promotion a decision you can defend, not a leap of faith.
The harness is the product. The autonomy ladder is how you adopt it, and the audit trail is what lets you climb.
Pre-launch · first design partners
Find the harness autonomy tier that fits your first workflow.
A Discovery engagement maps your highest-value, highest-scrutiny workflow to the HAT that fits, and proves the audit substrate against it before anything scales.
Become a design partner →